Please Note :-
Slides: will navigate with the right arrow >. Contains only main headers of the Project.
Sub-Slide: will navigate with the down arrow V. To see all the sub-slides click on V arrow untill its turned blank.
EDA Assignment - Lending Club Case Study Umesh Wagharalkar
Project Objectives
This company is the largest online loan marketplace, facilitating personal loans, business loans, and financing of medical procedures. Borrowers can easily access lower interest rate loans through a fast online interface.
Like most other lending companies, lending loans to ‘risky’ applicants is the largest source of financial loss (called credit loss). Credit loss is the amount of money lost by the lender when the borrower refuses to pay or runs away with the money owed. In other words, borrowers who default cause the largest amount of loss to the lenders. In this case, the customers labelled as 'charged-off' are the 'defaulters'.
If one is able to identify these risky loan applicants, then such loans can be reduced thereby cutting down the amount of credit loss. Identification of such applicants using EDA is the aim of this case study.
In other words, the company wants to understand the driving factors (or driver variables) behind loan default, i.e. the variables which are strong indicators of default. The company can utilise this knowledge for its portfolio and risk assessment.
Importing Libraries
In [422]:
# Importing all libraries required.
import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns
# set it to convert scientific values
pd.options.display.float_format = '{:.2f}'.format
# Regular expressions for string manipulation
import re
# Ignore warnings due to version problems
import warnings
warnings.filterwarnings('ignore')
Loading the data
In [490]:
data = pd.read_csv("loan.csv")
data.head(2)
Out[490]:
| id | member_id | loan_amnt | funded_amnt | funded_amnt_inv | term | int_rate | installment | grade | sub_grade | emp_title | emp_length | home_ownership | annual_inc | verification_status | issue_d | loan_status | pymnt_plan | url | desc | purpose | title | zip_code | addr_state | dti | delinq_2yrs | earliest_cr_line | inq_last_6mths | mths_since_last_delinq | mths_since_last_record | open_acc | pub_rec | revol_bal | revol_util | total_acc | initial_list_status | out_prncp | out_prncp_inv | total_pymnt | total_pymnt_inv | total_rec_prncp | total_rec_int | total_rec_late_fee | recoveries | collection_recovery_fee | last_pymnt_d | last_pymnt_amnt | next_pymnt_d | last_credit_pull_d | collections_12_mths_ex_med | mths_since_last_major_derog | policy_code | application_type | annual_inc_joint | dti_joint | verification_status_joint | acc_now_delinq | tot_coll_amt | tot_cur_bal | open_acc_6m | open_il_6m | open_il_12m | open_il_24m | mths_since_rcnt_il | total_bal_il | il_util | open_rv_12m | open_rv_24m | max_bal_bc | all_util | total_rev_hi_lim | inq_fi | total_cu_tl | inq_last_12m | acc_open_past_24mths | avg_cur_bal | bc_open_to_buy | bc_util | chargeoff_within_12_mths | delinq_amnt | mo_sin_old_il_acct | mo_sin_old_rev_tl_op | mo_sin_rcnt_rev_tl_op | mo_sin_rcnt_tl | mort_acc | mths_since_recent_bc | mths_since_recent_bc_dlq | mths_since_recent_inq | mths_since_recent_revol_delinq | num_accts_ever_120_pd | num_actv_bc_tl | num_actv_rev_tl | num_bc_sats | num_bc_tl | num_il_tl | num_op_rev_tl | num_rev_accts | num_rev_tl_bal_gt_0 | num_sats | num_tl_120dpd_2m | num_tl_30dpd | num_tl_90g_dpd_24m | num_tl_op_past_12m | pct_tl_nvr_dlq | percent_bc_gt_75 | pub_rec_bankruptcies | tax_liens | tot_hi_cred_lim | total_bal_ex_mort | total_bc_limit | total_il_high_credit_limit | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1077501 | 1296599 | 5000 | 5000 | 4975.00 | 36 months | 10.65% | 162.87 | B | B2 | NaN | 10+ years | RENT | 24000.00 | Verified | Dec-11 | Fully Paid | n | https://lendingclub.com/browse/loanDetail.acti... | Borrower added on 12/22/11 > I need to upgra... | credit_card | Computer | 860xx | AZ | 27.65 | 0 | Jan-85 | 1 | NaN | NaN | 3 | 0 | 13648 | 83.70% | 9 | f | 0.00 | 0.00 | 5863.16 | 5833.84 | 5000.00 | 863.16 | 0.00 | 0.00 | 0.00 | Jan-15 | 171.62 | NaN | May-16 | 0.00 | NaN | 1 | INDIVIDUAL | NaN | NaN | NaN | 0 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN |
| 1 | 1077430 | 1314167 | 2500 | 2500 | 2500.00 | 60 months | 15.27% | 59.83 | C | C4 | Ryder | < 1 year | RENT | 30000.00 | Source Verified | Dec-11 | Charged Off | n | https://lendingclub.com/browse/loanDetail.acti... | Borrower added on 12/22/11 > I plan to use t... | car | bike | 309xx | GA | 1.00 | 0 | Apr-99 | 5 | NaN | NaN | 3 | 0 | 1687 | 9.40% | 4 | f | 0.00 | 0.00 | 1008.71 | 1008.71 | 456.46 | 435.17 | 0.00 | 117.08 | 1.11 | Apr-13 | 119.66 | NaN | Sep-13 | 0.00 | NaN | 1 | INDIVIDUAL | NaN | NaN | NaN | 0 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN |
Inspecting The Dataframe
- Shape
- Info
- Data Types
- Describe
In [424]:
data.shape
Out[424]:
(39717, 111)
In [425]:
data.info()
<class 'pandas.core.frame.DataFrame'> RangeIndex: 39717 entries, 0 to 39716 Columns: 111 entries, id to total_il_high_credit_limit dtypes: float64(74), int64(13), object(24) memory usage: 33.6+ MB
In [426]:
data.dtypes
Out[426]:
id int64 member_id int64 loan_amnt int64 funded_amnt int64 funded_amnt_inv float64 term object int_rate object installment float64 grade object sub_grade object emp_title object emp_length object home_ownership object annual_inc float64 verification_status object issue_d object loan_status object pymnt_plan object url object desc object purpose object title object zip_code object addr_state object dti float64 delinq_2yrs int64 earliest_cr_line object inq_last_6mths int64 mths_since_last_delinq float64 mths_since_last_record float64 open_acc int64 pub_rec int64 revol_bal int64 revol_util object total_acc int64 initial_list_status object out_prncp float64 out_prncp_inv float64 total_pymnt float64 total_pymnt_inv float64 total_rec_prncp float64 total_rec_int float64 total_rec_late_fee float64 recoveries float64 collection_recovery_fee float64 last_pymnt_d object last_pymnt_amnt float64 next_pymnt_d object last_credit_pull_d object collections_12_mths_ex_med float64 mths_since_last_major_derog float64 policy_code int64 application_type object annual_inc_joint float64 dti_joint float64 verification_status_joint float64 acc_now_delinq int64 tot_coll_amt float64 tot_cur_bal float64 open_acc_6m float64 open_il_6m float64 open_il_12m float64 open_il_24m float64 mths_since_rcnt_il float64 total_bal_il float64 il_util float64 open_rv_12m float64 open_rv_24m float64 max_bal_bc float64 all_util float64 total_rev_hi_lim float64 inq_fi float64 total_cu_tl float64 inq_last_12m float64 acc_open_past_24mths float64 avg_cur_bal float64 bc_open_to_buy float64 bc_util float64 chargeoff_within_12_mths float64 delinq_amnt int64 mo_sin_old_il_acct float64 mo_sin_old_rev_tl_op float64 mo_sin_rcnt_rev_tl_op float64 mo_sin_rcnt_tl float64 mort_acc float64 mths_since_recent_bc float64 mths_since_recent_bc_dlq float64 mths_since_recent_inq float64 mths_since_recent_revol_delinq float64 num_accts_ever_120_pd float64 num_actv_bc_tl float64 num_actv_rev_tl float64 num_bc_sats float64 num_bc_tl float64 num_il_tl float64 num_op_rev_tl float64 num_rev_accts float64 num_rev_tl_bal_gt_0 float64 num_sats float64 num_tl_120dpd_2m float64 num_tl_30dpd float64 num_tl_90g_dpd_24m float64 num_tl_op_past_12m float64 pct_tl_nvr_dlq float64 percent_bc_gt_75 float64 pub_rec_bankruptcies float64 tax_liens float64 tot_hi_cred_lim float64 total_bal_ex_mort float64 total_bc_limit float64 total_il_high_credit_limit float64 dtype: object
In [427]:
data.describe()
Out[427]:
| id | member_id | loan_amnt | funded_amnt | funded_amnt_inv | installment | annual_inc | dti | delinq_2yrs | inq_last_6mths | mths_since_last_delinq | mths_since_last_record | open_acc | pub_rec | revol_bal | total_acc | out_prncp | out_prncp_inv | total_pymnt | total_pymnt_inv | total_rec_prncp | total_rec_int | total_rec_late_fee | recoveries | collection_recovery_fee | last_pymnt_amnt | collections_12_mths_ex_med | mths_since_last_major_derog | policy_code | annual_inc_joint | dti_joint | verification_status_joint | acc_now_delinq | tot_coll_amt | tot_cur_bal | open_acc_6m | open_il_6m | open_il_12m | open_il_24m | mths_since_rcnt_il | total_bal_il | il_util | open_rv_12m | open_rv_24m | max_bal_bc | all_util | total_rev_hi_lim | inq_fi | total_cu_tl | inq_last_12m | acc_open_past_24mths | avg_cur_bal | bc_open_to_buy | bc_util | chargeoff_within_12_mths | delinq_amnt | mo_sin_old_il_acct | mo_sin_old_rev_tl_op | mo_sin_rcnt_rev_tl_op | mo_sin_rcnt_tl | mort_acc | mths_since_recent_bc | mths_since_recent_bc_dlq | mths_since_recent_inq | mths_since_recent_revol_delinq | num_accts_ever_120_pd | num_actv_bc_tl | num_actv_rev_tl | num_bc_sats | num_bc_tl | num_il_tl | num_op_rev_tl | num_rev_accts | num_rev_tl_bal_gt_0 | num_sats | num_tl_120dpd_2m | num_tl_30dpd | num_tl_90g_dpd_24m | num_tl_op_past_12m | pct_tl_nvr_dlq | percent_bc_gt_75 | pub_rec_bankruptcies | tax_liens | tot_hi_cred_lim | total_bal_ex_mort | total_bc_limit | total_il_high_credit_limit | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| count | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 14035.00 | 2786.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39717.00 | 39661.00 | 0.00 | 39717.00 | 0.00 | 0.00 | 0.00 | 39717.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 39661.00 | 39717.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 39020.00 | 39678.00 | 0.00 | 0.00 | 0.00 | 0.00 |
| mean | 683131.91 | 850463.56 | 11219.44 | 10947.71 | 10397.45 | 324.56 | 68968.93 | 13.32 | 0.15 | 0.87 | 35.90 | 69.70 | 9.29 | 0.06 | 13382.53 | 22.09 | 51.23 | 50.99 | 12153.60 | 11567.15 | 9793.35 | 2263.66 | 1.36 | 95.22 | 12.41 | 2678.83 | 0.00 | NaN | 1.00 | NaN | NaN | NaN | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.04 | 0.00 | NaN | NaN | NaN | NaN |
| std | 210694.13 | 265678.31 | 7456.67 | 7187.24 | 7128.45 | 208.87 | 63793.77 | 6.68 | 0.49 | 1.07 | 22.02 | 43.82 | 4.40 | 0.24 | 15885.02 | 11.40 | 375.17 | 373.82 | 9042.04 | 8942.67 | 7065.52 | 2608.11 | 7.29 | 688.74 | 148.67 | 4447.14 | 0.00 | NaN | 0.00 | NaN | NaN | NaN | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.20 | 0.00 | NaN | NaN | NaN | NaN |
| min | 54734.00 | 70699.00 | 500.00 | 500.00 | 0.00 | 15.69 | 4000.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 2.00 | 0.00 | 0.00 | 2.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | 0.00 | NaN | 1.00 | NaN | NaN | NaN | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN |
| 25% | 516221.00 | 666780.00 | 5500.00 | 5400.00 | 5000.00 | 167.02 | 40404.00 | 8.17 | 0.00 | 0.00 | 18.00 | 22.00 | 6.00 | 0.00 | 3703.00 | 13.00 | 0.00 | 0.00 | 5576.93 | 5112.31 | 4600.00 | 662.18 | 0.00 | 0.00 | 0.00 | 218.68 | 0.00 | NaN | 1.00 | NaN | NaN | NaN | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN |
| 50% | 665665.00 | 850812.00 | 10000.00 | 9600.00 | 8975.00 | 280.22 | 59000.00 | 13.40 | 0.00 | 1.00 | 34.00 | 90.00 | 9.00 | 0.00 | 8850.00 | 20.00 | 0.00 | 0.00 | 9899.64 | 9287.15 | 8000.00 | 1348.91 | 0.00 | 0.00 | 0.00 | 546.14 | 0.00 | NaN | 1.00 | NaN | NaN | NaN | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN |
| 75% | 837755.00 | 1047339.00 | 15000.00 | 15000.00 | 14400.00 | 430.78 | 82300.00 | 18.60 | 0.00 | 1.00 | 52.00 | 104.00 | 12.00 | 0.00 | 17058.00 | 29.00 | 0.00 | 0.00 | 16534.43 | 15798.81 | 13653.26 | 2833.40 | 0.00 | 0.00 | 0.00 | 3293.16 | 0.00 | NaN | 1.00 | NaN | NaN | NaN | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN |
| max | 1077501.00 | 1314167.00 | 35000.00 | 35000.00 | 35000.00 | 1305.19 | 6000000.00 | 29.99 | 11.00 | 8.00 | 120.00 | 129.00 | 44.00 | 4.00 | 149588.00 | 90.00 | 6311.47 | 6307.37 | 58563.68 | 58563.68 | 35000.02 | 23563.68 | 180.20 | 29623.35 | 7002.19 | 36115.20 | 0.00 | NaN | 1.00 | NaN | NaN | NaN | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 0.00 | 0.00 | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | NaN | 2.00 | 0.00 | NaN | NaN | NaN | NaN |
Data Cleaning
In [428]:
# sum it up to check how many rows have all missing values
print('Number of empty Rows:')
print(data.isnull().all(axis=1).sum())
# Observation: There are no empty rows in whole dataset.
# sum it up to check how many columns have all missing values
print('Number of empty Columns:')
print(data.isnull().all(axis=0).sum())
# Observation: There are 54 columns have all missing values.
Number of empty Rows: 0 Number of empty Columns: 54
In [429]:
# Finding Which columns have all null values
data.isnull().sum()
Out[429]:
id 0 member_id 0 loan_amnt 0 funded_amnt 0 funded_amnt_inv 0 term 0 int_rate 0 installment 0 grade 0 sub_grade 0 emp_title 2459 emp_length 1075 home_ownership 0 annual_inc 0 verification_status 0 issue_d 0 loan_status 0 pymnt_plan 0 url 0 desc 12940 purpose 0 title 11 zip_code 0 addr_state 0 dti 0 delinq_2yrs 0 earliest_cr_line 0 inq_last_6mths 0 mths_since_last_delinq 25682 mths_since_last_record 36931 open_acc 0 pub_rec 0 revol_bal 0 revol_util 50 total_acc 0 initial_list_status 0 out_prncp 0 out_prncp_inv 0 total_pymnt 0 total_pymnt_inv 0 total_rec_prncp 0 total_rec_int 0 total_rec_late_fee 0 recoveries 0 collection_recovery_fee 0 last_pymnt_d 71 last_pymnt_amnt 0 next_pymnt_d 38577 last_credit_pull_d 2 collections_12_mths_ex_med 56 mths_since_last_major_derog 39717 policy_code 0 application_type 0 annual_inc_joint 39717 dti_joint 39717 verification_status_joint 39717 acc_now_delinq 0 tot_coll_amt 39717 tot_cur_bal 39717 open_acc_6m 39717 open_il_6m 39717 open_il_12m 39717 open_il_24m 39717 mths_since_rcnt_il 39717 total_bal_il 39717 il_util 39717 open_rv_12m 39717 open_rv_24m 39717 max_bal_bc 39717 all_util 39717 total_rev_hi_lim 39717 inq_fi 39717 total_cu_tl 39717 inq_last_12m 39717 acc_open_past_24mths 39717 avg_cur_bal 39717 bc_open_to_buy 39717 bc_util 39717 chargeoff_within_12_mths 56 delinq_amnt 0 mo_sin_old_il_acct 39717 mo_sin_old_rev_tl_op 39717 mo_sin_rcnt_rev_tl_op 39717 mo_sin_rcnt_tl 39717 mort_acc 39717 mths_since_recent_bc 39717 mths_since_recent_bc_dlq 39717 mths_since_recent_inq 39717 mths_since_recent_revol_delinq 39717 num_accts_ever_120_pd 39717 num_actv_bc_tl 39717 num_actv_rev_tl 39717 num_bc_sats 39717 num_bc_tl 39717 num_il_tl 39717 num_op_rev_tl 39717 num_rev_accts 39717 num_rev_tl_bal_gt_0 39717 num_sats 39717 num_tl_120dpd_2m 39717 num_tl_30dpd 39717 num_tl_90g_dpd_24m 39717 num_tl_op_past_12m 39717 pct_tl_nvr_dlq 39717 percent_bc_gt_75 39717 pub_rec_bankruptcies 697 tax_liens 39 tot_hi_cred_lim 39717 total_bal_ex_mort 39717 total_bc_limit 39717 total_il_high_credit_limit 39717 dtype: int64
In [430]:
# Droping columns where all values are null. Below in sequence all columns have null values
data.dropna(how='all', inplace = True, axis = 1)
In [431]:
# Droping additional columns which are not needed for the analysis.
columns_drop = ['delinq_2yrs','earliest_cr_line','inq_last_6mths','open_acc','pub_rec','revol_bal','revol_util','total_acc','out_prncp','out_prncp_inv','total_pymnt','total_pymnt_inv','total_rec_prncp','total_rec_int','total_rec_late_fee','recoveries','collection_recovery_fee','last_pymnt_d','last_pymnt_amnt','last_credit_pull_d','application_type','policy_code','initial_list_status','installment','pymnt_plan','mths_since_last_delinq','mths_since_last_record','next_pymnt_d','desc','emp_title']
data.drop(labels = columns_drop, axis =1, inplace=True)
print(data.shape)
data.isnull().sum()
(39717, 27)
Out[431]:
id 0 member_id 0 loan_amnt 0 funded_amnt 0 funded_amnt_inv 0 term 0 int_rate 0 grade 0 sub_grade 0 emp_length 1075 home_ownership 0 annual_inc 0 verification_status 0 issue_d 0 loan_status 0 url 0 purpose 0 title 11 zip_code 0 addr_state 0 dti 0 collections_12_mths_ex_med 56 acc_now_delinq 0 chargeoff_within_12_mths 56 delinq_amnt 0 pub_rec_bankruptcies 697 tax_liens 39 dtype: int64
In [432]:
# Data shape after dropping columns which had nothing
data.shape
# Observations
# So now we are left with 38 columns
Out[432]:
(39717, 27)
In [433]:
# Formating emp_length column to have only numbers.
data.emp_length.fillna('0',inplace=True)
data['emp_length']=data.emp_length.str.extract('(\d+)')
data.head(2)
Out[433]:
| id | member_id | loan_amnt | funded_amnt | funded_amnt_inv | term | int_rate | grade | sub_grade | emp_length | home_ownership | annual_inc | verification_status | issue_d | loan_status | url | purpose | title | zip_code | addr_state | dti | collections_12_mths_ex_med | acc_now_delinq | chargeoff_within_12_mths | delinq_amnt | pub_rec_bankruptcies | tax_liens | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1077501 | 1296599 | 5000 | 5000 | 4975.00 | 36 months | 10.65% | B | B2 | 10 | RENT | 24000.00 | Verified | Dec-11 | Fully Paid | https://lendingclub.com/browse/loanDetail.acti... | credit_card | Computer | 860xx | AZ | 27.65 | 0.00 | 0 | 0.00 | 0 | 0.00 | 0.00 |
| 1 | 1077430 | 1314167 | 2500 | 2500 | 2500.00 | 60 months | 15.27% | C | C4 | 1 | RENT | 30000.00 | Source Verified | Dec-11 | Charged Off | https://lendingclub.com/browse/loanDetail.acti... | car | bike | 309xx | GA | 1.00 | 0.00 | 0 | 0.00 | 0 | 0.00 | 0.00 |
In [434]:
# pub_rec_bankruptcies can be cleaned for further analysis.
print("Before cleaning:")
print(data.pub_rec_bankruptcies.isnull().sum())
print("After cleaning:")
data.pub_rec_bankruptcies.fillna('Not Known',inplace=True)
print(data.pub_rec_bankruptcies.isnull().sum())
Before cleaning: 697 After cleaning: 0
In [435]:
# Removeing % symbol from int_rate column
data['int_rate'] = data['int_rate'].str.rstrip('%')
In [436]:
# convert amount columns into numeric data to find some correlation.
cols1 = ['loan_amnt','funded_amnt','int_rate','funded_amnt_inv','annual_inc','dti','emp_length']
data[cols1] = data[cols1].apply(pd.to_numeric)
In [437]:
# Lets take a look on important variables loan status in total loan issued
(data.loan_status.value_counts()*100)/len(data)
# Observations
# 82.96 percent loans were fully paid.
# Close to 14% loans were charged off
Out[437]:
Fully Paid 82.96 Charged Off 14.17 Current 2.87 Name: loan_status, dtype: float64
In [438]:
# Lets take a look on important variable purpose of loan in total loans issued
(data.purpose.value_counts()*100)/len(data)
# Observations
# Most of the loans taken for debt consolidation(47%) and Credit card bill payment
# Other is also one of the main area where loans were taken.
Out[438]:
debt_consolidation 46.93 credit_card 12.92 other 10.05 home_improvement 7.49 major_purchase 5.51 small_business 4.60 car 3.90 wedding 2.38 medical 1.74 moving 1.47 vacation 0.96 house 0.96 educational 0.82 renewable_energy 0.26 Name: purpose, dtype: float64
Derive Columns For Analysis -
In [439]:
# Lets derive some new columns using existing data as it may further used in analysis.
# create month and year columns separately
data.issue_d = pd.to_datetime(data.issue_d, format='%b-%y')
data['year']=data['issue_d'].dt.year
data['month']=data['issue_d'].dt.month
data.head(1)
# Observations
# Two new columns month and year were derived from issue date
Out[439]:
| id | member_id | loan_amnt | funded_amnt | funded_amnt_inv | term | int_rate | grade | sub_grade | emp_length | home_ownership | annual_inc | verification_status | issue_d | loan_status | url | purpose | title | zip_code | addr_state | dti | collections_12_mths_ex_med | acc_now_delinq | chargeoff_within_12_mths | delinq_amnt | pub_rec_bankruptcies | tax_liens | year | month | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1077501 | 1296599 | 5000 | 5000 | 4975.00 | 36 months | 10.65 | B | B2 | 10 | RENT | 24000.00 | Verified | 2011-12-01 | Fully Paid | https://lendingclub.com/browse/loanDetail.acti... | credit_card | Computer | 860xx | AZ | 27.65 | 0.00 | 0 | 0.00 | 0 | 0.00 | 0.00 | 2011 | 12 |
In [440]:
# categorise loan amounts into buckets which will help in analysis further in bivariate analysis.
data['loan_amnt_cats'] = pd.cut(data['loan_amnt'], [0, 7000, 14000, 21000, 28000, 35000], labels=['0-7000', '7000-14000', '14000-21000', '21000-28000', '28000 +'])
# categorise annual incomes into buckets which will help in analysis further in bivariate analysis.
data['annual_inc_cats'] = pd.cut(data['annual_inc'], [0, 20000, 40000, 60000, 80000,1000000], labels=['0-20000', '20000-40000', '40000-60000', '60000-80000', '80000 +'])
# categorise intrest rates into buckets which will help in analysis further in bivariate analysis.
data['int_rate_cats'] = pd.cut(data['int_rate'], [0, 10, 12.5, 16, 20], labels=['0-10', '10-13', '12.5-16', '16 +'])
# categorise dti into buckets for bivariate analysis.
data['dti_cats'] = pd.cut(data['dti'], [0, 5, 10, 15, 20, 25], labels=['0-5', '05-10', '10-15', '15-20', '25+'])
In [441]:
# Lets take a look on shape and data
print(data.shape)
data.head(1)
# Observation is we have created extra 4 columns after dropping many.
(39717, 33)
Out[441]:
| id | member_id | loan_amnt | funded_amnt | funded_amnt_inv | term | int_rate | grade | sub_grade | emp_length | home_ownership | annual_inc | verification_status | issue_d | loan_status | url | purpose | title | zip_code | addr_state | dti | collections_12_mths_ex_med | acc_now_delinq | chargeoff_within_12_mths | delinq_amnt | pub_rec_bankruptcies | tax_liens | year | month | loan_amnt_cats | annual_inc_cats | int_rate_cats | dti_cats | |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 0 | 1077501 | 1296599 | 5000 | 5000 | 4975.00 | 36 months | 10.65 | B | B2 | 10 | RENT | 24000.00 | Verified | 2011-12-01 | Fully Paid | https://lendingclub.com/browse/loanDetail.acti... | credit_card | Computer | 860xx | AZ | 27.65 | 0.00 | 0 | 0.00 | 0 | 0.00 | 0.00 | 2011 | 12 | 0-7000 | 20000-40000 | 10-13 | NaN |
Univariate Analysis Part 1 -
In [442]:
# Basic statistics with .describe() - Quantitative Variables
data['loan_amnt'].describe()
Out[442]:
count 39717.00 mean 11219.44 std 7456.67 min 500.00 25% 5500.00 50% 10000.00 75% 15000.00 max 35000.00 Name: loan_amnt, dtype: float64
In [443]:
sns.boxplot(data.loan_amnt)
Out[443]:
<AxesSubplot:xlabel='loan_amnt'>
In [444]:
# Basic statistics with .describe() -Quantitative Variables
print('Before Removal of Outliers :\n')
print(data['annual_inc'].describe())
# Removing Outliers quantile .99 from Annual Income
# it will make it easier to visualize the plots.
data = data[data["annual_inc"] < data["annual_inc"].quantile(0.99)]
print('After Removal of Outliers :')
print(data["annual_inc"].describe())
# Now below data looks much better. Lets plot later and find some conclusions
Before Removal of Outliers : count 39717.00 mean 68968.93 std 63793.77 min 4000.00 25% 40404.00 50% 59000.00 75% 82300.00 max 6000000.00 Name: annual_inc, dtype: float64 After Removal of Outliers : count 39319.00 mean 65524.22 std 35215.89 min 4000.00 25% 40000.00 50% 58000.00 75% 81000.00 max 234996.00 Name: annual_inc, dtype: float64
In [445]:
sns.boxplot(data.annual_inc)
Out[445]:
<AxesSubplot:xlabel='annual_inc'>
In [446]:
# Basic statistics with .describe() -Quantitative Variables
data['int_rate'].describe()
# Average intrest rate is 12 %
# after 75% percentile interest rate zoom'ed to 25% from 15%.
Out[446]:
count 39319.00 mean 12.01 std 3.72 min 5.42 25% 9.25 50% 11.83 75% 14.54 max 24.59 Name: int_rate, dtype: float64
In [447]:
sns.boxplot(data.int_rate)
Out[447]:
<AxesSubplot:xlabel='int_rate'>
Univariate Analysis Part 2 -
In [448]:
# Lets see distribution of three loan amount fields using distribution plot.
# Quantitative Variables
plt.figure(figsize=(15,8),facecolor='b')
sns.set_style("dark")
# subplot 1
plt.subplot(2, 3, 1)
ax = sns.distplot(data['loan_amnt'],rug = True)
ax.set_title('Loan Amount - Distribution Plot',fontsize=14,color='w')
ax.set_xlabel('Loan Amount',fontsize=14,color='w')
# subplot 2
plt.subplot(2, 3, 2)
ax = sns.distplot(data['funded_amnt'],rug = True)
ax.set_title('Funded Amount - Distribution Plot',fontsize=14,color='w')
ax.set_xlabel('Funded Amount',fontsize=14,color='w')
# subplot 2
plt.subplot(2, 3, 3)
ax = sns.distplot(data['funded_amnt_inv'],rug = True)
ax.set_title('Funded Amount Inv. - Distribution Plot',fontsize=14,color='w')
ax.set_xlabel('Funded Amount Inv.',fontsize=14,color='w')
plt.show()
# Observation:
# Distribution of amounts for all three looks very much similar.
# We will work with only loan amount column for rest of our analysis.
In [449]:
# Univariate Analysis on Loan amount-Quantitative Variables
plt.figure(figsize=(15,8),facecolor='b')
sns.set_style("dark")
# subplot 1
plt.subplot(2, 2, 1)
ax = sns.distplot(data['loan_amnt'],rug = True)
ax.set_title('Loan Amount - Distribution Plot',fontsize=16,color='w')
ax.set_xlabel('Loan Amount',fontsize=14,color='w')
# subplot 2
plt.subplot(2, 2, 2)
ax = sns.boxplot(y=data['loan_amnt'])
ax.set_title('Loan Amount - Box Plot',fontsize=16,color='w')
ax.set_ylabel('Loan Amount',fontsize=14,color='w')
plt.show()
# Observations :
# Below plots show that most of the Loan amounts are in range of 5000 - 15000
In [450]:
# Univariate Analysis on Intrest Rate-Quantitative Variables
plt.figure(figsize=(15,8),facecolor='b')
sns.set_style("dark")
# subplot 1
plt.subplot(2, 2, 1)
ax = sns.distplot(data['int_rate'],rug = True)
ax.set_title('Interest Rate - Distribution Plot',fontsize=16,color='w')
ax.set_xlabel('Interest Rate',fontsize=14,color='w')
# subplot 2
plt.subplot(2, 2, 2)
ax = sns.boxplot(y=data['int_rate'])
ax.set_title('Interest Rate - Box Plot',fontsize=16,color='w')
ax.set_ylabel('Interest Rate',fontsize=14,color='w')
plt.show()
# Observations :
# Below plots show that most of the Interest Rates on loans are in range of 10% - 15%
In [451]:
# Univariate Analysis on Annual Income - Quantitative Variables
plt.figure(figsize=(15,8),facecolor='b')
sns.set_style("dark")
# subplot 1
plt.subplot(2, 2, 1)
ax = sns.distplot(data['annual_inc'],rug = True)
ax.set_title('Annual Income - Distribution Plot',fontsize=16,color='w')
ax.set_xlabel('Annual Income',fontsize=14,color='w')
# subplot 2
plt.subplot(2, 2, 2)
plt.title('Annual Income - Box Plot')
ax = sns.boxplot(y=data['annual_inc'])
ax.set_title('Annual Income - Box Plot',fontsize=16,color='w')
ax.set_ylabel('Annual Income',fontsize=14,color='w')
plt.show()
# Observations :
# Below plots show that most of the borrower's Annual incomes are in range of 40000- 80000
In [452]:
# Univariate Analysis - Unordered Categorical Variables - Loan Status
plt.figure(figsize=(10,8),facecolor='g')
sns.set_style("dark")
ax = sns.countplot(x="loan_status",data=data,palette='mako')
ax.set_title('Loan Status',fontsize=14,color='w')
ax.set_xlabel('Loan Status',fontsize=14,color = 'w')
ax.set_ylabel('Loan Application Count',fontsize=14,color = 'w')
# To show count of values above bars
s=data['loan_status'].value_counts()
for i, v in s.reset_index().iterrows():
ax.text(i, v.loan_status + 0.3 , v.loan_status, color='k')
# Observations :
# Below plot shows that close to 14% loans were charged off out of total loan issued.
In [454]:
# Univariate Analysis - Unordered Categorical Variables - Purpose Of Loan
plt.figure(figsize=(14,8),facecolor='g')
sns.set_style("dark")
ax = sns.countplot(y="purpose",data=data,hue='loan_status',palette='GnBu_d')
ax.set_title('Purpose Of Loan',fontsize=14,color='w')
ax.set_ylabel('Purpose Of Loan',fontsize=14,color = 'w')
ax.set_xlabel('Loan Application Count',fontsize=14,color = 'w')
plt.show()
# Observations :
# Below plot shows that most of the loans were taken for the purpose of debt consolidation & paying credit card bill.
# Number of chraged off count also high too for these loans.
In [455]:
# Univariate Analysis - Unordered Categorical Variables - Home Ownership
plt.figure(figsize=(10,6),facecolor='g')
ax = sns.countplot(x="home_ownership",data=data,hue='loan_status',palette='GnBu_d')
ax.legend(bbox_to_anchor=(1, 1))
ax.set_title('Home Ownership',fontsize=14,color='w')
ax.set_xlabel('Home Ownership',fontsize=14,color = 'w')
ax.set_ylabel('Loan Application Count',fontsize=14,color = 'w')
plt.show()
# Observations :
# Below plot shows that most of them living in rented home or mortgazed their home.
# Applicant numbers are high from these categories so charged off is high too.
In [456]:
# Derived Column - Ordered Categorical Variables
# Let us look into number of loans which were approved every year/month
# Lets use derived column year to check pattern of loan issuing over the years.
plt.figure(figsize=(14,6),facecolor='g')
data.groupby(['year','month']).id.count().plot(kind='bar')
plt.ylabel('Loan applications count',fontsize=14,color='w')
plt.xlabel('Year,Month',fontsize=14,color = 'w')
plt.title("No of loan applications issued over the years",fontsize=14,color='w')
plt.show()
# Observation is that count of loan application is increasing every passing year.
# so increase in number of loan applications are adding more to number of charged off applications.
# number of loans issued in 2008( May-October) got dipped, may be due to Recession.
In [457]:
# Univariate Analysis - Ordered Categorical Variables- Loan Paying Term
plt.figure(figsize=(10,6),facecolor='g')
ax = sns.countplot(x="term",data=data,hue='loan_status',palette='GnBu_d')
ax.set_title('Loan Paying Term',fontsize=14,color='w')
ax.set_xlabel('Loan Repayment Term',fontsize=14,color = 'w')
ax.set_ylabel('Loan Application Count',fontsize=14,color = 'w')
ax.legend(bbox_to_anchor=(1, 1))
plt.show()
# Observations :
# Below plot shows that those who had taken loan to repay in 60 months had more % of number of applicants getting
# charged off as compared to applicants who had taken loan for 36 months.
Correlation Matrix - Quantitative Variables
In [458]:
# Bivariate Analysis - Correlation Matrix-Quantitative Variables
# Heatmap with Dendogram(clustermap) to show closeness among numerical variables.
loan_correlation = A.corr()
sns.set(font_scale=1.1)
sns.clustermap(loan_correlation,annot=True,figsize=(12, 8),cmap="BrBG")
plt.show()
# Observation is that Loan amount, investor amount, funding amount are strongly correlated.
# Annual income with DTI(Debt-to-income ratio) is negatively correalted.
# Debt income ratio is the percentage of a consumer's monthly gross income that goes toward paying debts.
# That means when annual income is low DTI is high & vice versa.
# positive correlation between annual income and employment years.
# That means income increases with work experience ;)
Bivariate Analysis : Part 1
In [459]:
# Bivariate Analysis on annual income against Chargedoff_Proportion -
inc_range_vs_loan = data.groupby(['annual_inc_cats', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
inc_range_vs_loan['Total'] = inc_range_vs_loan['Charged Off'] + inc_range_vs_loan['Current'] + inc_range_vs_loan['Fully Paid']
inc_range_vs_loan['Chargedoff_Proportion'] = inc_range_vs_loan['Charged Off'] / inc_range_vs_loan['Total']
inc_range_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# Income range 80000+ has less chances of charged off.
# Income range 0-20000 has high chances of charged off.
# Notice that with increase in annual income charged off proportion got decreased.
Out[459]:
| loan_status | annual_inc_cats | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 0 | 0-20000 | 237 | 9 | 943 | 1189 | 0.20 |
| 1 | 20000-40000 | 1514 | 170 | 7004 | 8688 | 0.17 |
| 2 | 40000-60000 | 1729 | 345 | 9534 | 11608 | 0.15 |
| 3 | 60000-80000 | 1024 | 240 | 6597 | 7861 | 0.13 |
| 4 | 80000 + | 1080 | 362 | 8531 | 9973 | 0.11 |
In [460]:
# Drawing bar plots on data calculated above. Try to visualize the pattern to understand the data better.
fig, ax1 = plt.subplots(figsize=(12, 8),facecolor='g')
ax1.set_title('Annual Income vs Chargedoff Proportion',fontsize=15,color = 'w')
ax1=sns.barplot(x='annual_inc_cats', y='Chargedoff_Proportion', data=inc_range_vs_loan)
ax1.set_ylabel('Chargedoff Proportion',fontsize=14,color = 'w')
ax1.set_xlabel('Annual Income',fontsize=14,color='w')
plt.show()
# Observations:
# Income range 80000+ has less chances of charged off.
# Income range 0-20000 has high chances of charged off.
# Notice that with increase in annual income charged off proportion got decreased.
In [461]:
# Bivariate Analysis on purpose of Loan against Chargedoff_Proportion.
purpose_vs_loan = data.groupby(['purpose', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
purpose_vs_loan['Total'] = purpose_vs_loan['Charged Off'] + purpose_vs_loan['Current'] + purpose_vs_loan['Fully Paid']
purpose_vs_loan['Chargedoff_Proportion'] = purpose_vs_loan['Charged Off'] / purpose_vs_loan['Total']
purpose_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# small Business applicants have high chnaces of getting charged off.
# renewable_energy where chanrged off proportion is better as compare to other categories.
Out[461]:
| loan_status | purpose | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 11 | small_business | 472.00 | 74.00 | 1251.00 | 1797.00 | 0.26 |
| 10 | renewable_energy | 19.00 | 1.00 | 81.00 | 101.00 | 0.19 |
| 3 | educational | 55.00 | 0.00 | 268.00 | 323.00 | 0.17 |
| 9 | other | 629.00 | 126.00 | 3203.00 | 3958.00 | 0.16 |
| 8 | moving | 91.00 | 7.00 | 479.00 | 577.00 | 0.16 |
| 5 | house | 58.00 | 14.00 | 299.00 | 371.00 | 0.16 |
| 7 | medical | 106.00 | 12.00 | 568.00 | 686.00 | 0.15 |
| 2 | debt_consolidation | 2752.00 | 580.00 | 15165.00 | 18497.00 | 0.15 |
| 12 | vacation | 53.00 | 6.00 | 322.00 | 381.00 | 0.14 |
| 4 | home_improvement | 338.00 | 97.00 | 2459.00 | 2894.00 | 0.12 |
| 1 | credit_card | 537.00 | 102.00 | 4450.00 | 5089.00 | 0.11 |
| 0 | car | 160.00 | 50.00 | 1332.00 | 1542.00 | 0.10 |
| 6 | major_purchase | 220.00 | 37.00 | 1909.00 | 2166.00 | 0.10 |
| 13 | wedding | 94.00 | 20.00 | 823.00 | 937.00 | 0.10 |
In [462]:
# Drowing bar plots on data calculated above. Try to visualize the pattern to understand the data better.
# Pairs of continuous variables.
fig, ax1 = plt.subplots(figsize=(14, 8),facecolor='g')
ax1.set_title('Purpose of Loans vs Chargedoff Proportion',fontsize=15,color = 'w')
ax1=sns.barplot(y='purpose', x='Chargedoff_Proportion', data=purpose_vs_loan)
ax1.set_ylabel('Purpose of Loans',fontsize=14,color='w')
ax1.set_xlabel('Chargedoff Proportion',fontsize=14,color = 'w')
plt.show()
# Observations:
# small Business applicants have high chances of getting charged off.
# renewable_energy where chanrged off proportion is better as compare to other categories.
In [463]:
# Bivariate Analysis on grade against Chargedoff_Proportion
grade_vs_loan = data.groupby(['grade', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
grade_vs_loan['Total'] = grade_vs_loan['Charged Off'] + grade_vs_loan['Current'] + grade_vs_loan['Fully Paid']
grade_vs_loan['Chargedoff_Proportion'] = grade_vs_loan['Charged Off'] / grade_vs_loan['Total']
grade_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# Grade "A" has very less chances of charged off.
# Grade "F" and "G" have very high chances of charged off.
# Chances of charged off is increasing with grade moving from "A" towards "G"
Out[463]:
| loan_status | grade | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 6 | G | 99 | 17 | 190 | 306 | 0.32 |
| 5 | F | 312 | 71 | 642 | 1025 | 0.30 |
| 4 | E | 704 | 176 | 1906 | 2786 | 0.25 |
| 3 | D | 1112 | 216 | 3916 | 5244 | 0.21 |
| 2 | C | 1342 | 262 | 6411 | 8015 | 0.17 |
| 1 | B | 1414 | 344 | 10167 | 11925 | 0.12 |
| 0 | A | 601 | 40 | 9377 | 10018 | 0.06 |
In [464]:
# Drawing bar plots on data calculated above. Try to visualize the pattern to understand the data better.
fig, ax1 = plt.subplots(figsize=(14, 8),facecolor='g')
ax1.set_title('Grades vs Chargedoff Proportion',fontsize=15,color='w')
ax1=sns.barplot(x='grade', y='Chargedoff_Proportion', data=grade_vs_loan)
ax1.set_xlabel('Grades',fontsize=14,color='w')
ax1.set_ylabel('Chargedoff Proportion',fontsize=14,color ='w')
plt.show()
# Observations:
# Grade "A" has very less chances of charged off.
# Grade "F" and "G" have very high chances of charged off.
# Chances of charged of is increasing with grade moving from "A" towards "G"
In [465]:
# Understanding grades better
# Bivariate Analysis on sub grade against Chargedoff_Proportion
grade_vs_loan = data.groupby(['sub_grade', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
grade_vs_loan['Total'] = grade_vs_loan['Charged Off'] + grade_vs_loan['Current'] + grade_vs_loan['Fully Paid']
grade_vs_loan['Chargedoff_Proportion'] = grade_vs_loan['Charged Off'] / grade_vs_loan['Total']
grade_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# sub Grades of "A" has very less chances of charged off.
# sub Grades of "F" and "G" have very high chances of charged off.
# proportion of charged off is increasing with sub grades moving from sub grades of "A" towards sub grades of "G"
Out[465]:
| loan_status | sub_grade | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 29 | F5 | 53.00 | 4.00 | 55.00 | 112.00 | 0.47 |
| 32 | G3 | 19.00 | 3.00 | 25.00 | 47.00 | 0.40 |
| 31 | G2 | 27.00 | 1.00 | 47.00 | 75.00 | 0.36 |
| 34 | G5 | 10.00 | 1.00 | 18.00 | 29.00 | 0.34 |
| 28 | F4 | 52.00 | 17.00 | 96.00 | 165.00 | 0.32 |
| 30 | G1 | 31.00 | 10.00 | 61.00 | 102.00 | 0.30 |
| 23 | E4 | 126.00 | 28.00 | 293.00 | 447.00 | 0.28 |
| 26 | F2 | 69.00 | 16.00 | 162.00 | 247.00 | 0.28 |
| 27 | F3 | 50.00 | 11.00 | 118.00 | 179.00 | 0.28 |
| 25 | F1 | 88.00 | 23.00 | 211.00 | 322.00 | 0.27 |
| 24 | E5 | 106.00 | 29.00 | 264.00 | 399.00 | 0.27 |
| 20 | E1 | 195.00 | 40.00 | 517.00 | 752.00 | 0.26 |
| 21 | E2 | 160.00 | 42.00 | 444.00 | 646.00 | 0.25 |
| 19 | D5 | 209.00 | 36.00 | 617.00 | 862.00 | 0.24 |
| 33 | G4 | 12.00 | 2.00 | 39.00 | 53.00 | 0.23 |
| 18 | D4 | 214.00 | 63.00 | 693.00 | 970.00 | 0.22 |
| 17 | D3 | 253.00 | 56.00 | 850.00 | 1159.00 | 0.22 |
| 22 | E3 | 117.00 | 37.00 | 388.00 | 542.00 | 0.22 |
| 16 | D2 | 269.00 | 61.00 | 1000.00 | 1330.00 | 0.20 |
| 15 | D1 | 167.00 | 0.00 | 756.00 | 923.00 | 0.18 |
| 12 | C3 | 269.00 | 41.00 | 1200.00 | 1510.00 | 0.18 |
| 14 | C5 | 208.00 | 32.00 | 931.00 | 1171.00 | 0.18 |
| 13 | C4 | 211.00 | 30.00 | 988.00 | 1229.00 | 0.17 |
| 11 | C2 | 321.00 | 78.00 | 1593.00 | 1992.00 | 0.16 |
| 10 | C1 | 333.00 | 81.00 | 1699.00 | 2113.00 | 0.16 |
| 9 | B5 | 356.00 | 89.00 | 2244.00 | 2689.00 | 0.13 |
| 8 | B4 | 328.00 | 74.00 | 2097.00 | 2499.00 | 0.13 |
| 7 | B3 | 335.00 | 92.00 | 2457.00 | 2884.00 | 0.12 |
| 6 | B2 | 225.00 | 56.00 | 1753.00 | 2034.00 | 0.11 |
| 5 | B1 | 170.00 | 33.00 | 1616.00 | 1819.00 | 0.09 |
| 4 | A5 | 216.00 | 27.00 | 2478.00 | 2721.00 | 0.08 |
| 3 | A4 | 178.00 | 13.00 | 2676.00 | 2867.00 | 0.06 |
| 2 | A3 | 103.00 | 0.00 | 1696.00 | 1799.00 | 0.06 |
| 1 | A2 | 74.00 | 0.00 | 1425.00 | 1499.00 | 0.05 |
| 0 | A1 | 30.00 | 0.00 | 1102.00 | 1132.00 | 0.03 |
In [466]:
# Drawing bar plots on data calculated above. Try to visualize the pattern to understand the data better.
fig, ax1 = plt.subplots(figsize=(14, 8),facecolor='g')
ax1.set_title('Sub Grades vs Chargedoff Proportion',fontsize=15,color='w')
ax1=sns.barplot(x='sub_grade', y='Chargedoff_Proportion', data=grade_vs_loan)
ax1.set_xlabel('Sub Grades',fontsize=14,color='w')
ax1.set_ylabel('Chargedoff Proportion',fontsize=14,color = 'w')
plt.show()
# Observations:
# sub Grades of "A" has very less chances of charged off.
# sub Grades of "F" and "G" have very high chances of charged off.
# proportion of charged off is increasing with sub grades moving from sub grades of "A" towards sub grades of "G"
In [467]:
# Bivariate Analysis on interest rate against Chargedoff_Proportion
interest_vs_loan = data.groupby(['int_rate_cats', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
interest_vs_loan['Total'] = interest_vs_loan['Charged Off'] + interest_vs_loan['Current'] + interest_vs_loan['Fully Paid']
interest_vs_loan['Chargedoff_Proportion'] = interest_vs_loan['Charged Off'] / interest_vs_loan['Total']
interest_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# interest rate less than 10% has very less chances of charged off. Intrest rates are starting from minimin 5 %.
# interest rate more than 16% has good chnaces of charged off as compared to other category intrest rates.
# Charged off proportion is increasing with higher intrest rates.
Out[467]:
| loan_status | int_rate_cats | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 3 | 16 + | 1250 | 351 | 3317 | 4918 | 0.25 |
| 2 | 12.5-16 | 1995 | 329 | 9354 | 11678 | 0.17 |
| 1 | 10-13 | 1224 | 269 | 8083 | 9576 | 0.13 |
| 0 | 0-10 | 825 | 77 | 11403 | 12305 | 0.07 |
In [468]:
# Drawing some bar plots on data calculated above. Try to visualize the pattern to understand the data better.
fig, ax1 = plt.subplots(figsize=(12, 8),facecolor='g')
ax1.set_title('Interest Rate vs Chargedoff Proportion',fontsize=15,color='w')
ax1=sns.barplot(x='int_rate_cats', y='Chargedoff_Proportion', data=interest_vs_loan)
ax1.set_xlabel('Interest Rate',fontsize=14,color='w')
ax1.set_ylabel('Chargedoff Proportion',fontsize=14,color = 'w')
plt.show()
# Observations:
# interest rate less than 10% has very less chances of charged off. Intrest rates are starting from minimin 5 %.
# interest rate more than 16% has good chnaces of charged off as compared to other category intrest rates.
# Charged off proportion is increasing with higher intrest rates.
In [469]:
# Bivariate Analysis on employment length against Chargedoff_Proportion
emplen_vs_loan = data.groupby(['emp_length', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
emplen_vs_loan['Total'] = emplen_vs_loan['Charged Off'] + emplen_vs_loan['Current'] + emplen_vs_loan['Fully Paid']
emplen_vs_loan['Chargedoff_Proportion'] = emplen_vs_loan['Charged Off'] / emplen_vs_loan['Total']
emplen_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# Those who are not working or have less than 1 year of work experience have high chances of getting charged off.
# It makes sense as with less or no experience they don't have source of income to repay loan.
# Rest of the applicants have more or less same chnaces of getting charged off.
Out[469]:
| loan_status | emp_length | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 0 | 0 | 227 | 42 | 801 | 1070 | 0.21 |
| 10 | 10 | 1316 | 386 | 7038 | 8740 | 0.15 |
| 7 | 7 | 262 | 62 | 1435 | 1759 | 0.15 |
| 1 | 1 | 1090 | 143 | 6533 | 7766 | 0.14 |
| 5 | 5 | 456 | 87 | 2712 | 3255 | 0.14 |
| 8 | 8 | 203 | 43 | 1216 | 1462 | 0.14 |
| 6 | 6 | 305 | 58 | 1846 | 2209 | 0.14 |
| 3 | 3 | 551 | 82 | 3426 | 4059 | 0.14 |
| 4 | 4 | 456 | 94 | 2860 | 3410 | 0.13 |
| 2 | 2 | 561 | 97 | 3684 | 4342 | 0.13 |
| 9 | 9 | 157 | 32 | 1058 | 1247 | 0.13 |
In [470]:
# Drawing bar plots on data calculated above. Try to visualize the pattern to understand the data better.
fig, ax1 = plt.subplots(figsize=(12, 8),facecolor='g')
ax1.set_title('Employment Length vs Chargedoff Proportion',fontsize=15,color='w')
ax1=sns.barplot(x='emp_length', y='Chargedoff_Proportion', data=emplen_vs_loan)
ax1.set_xlabel('Employment Length',fontsize=14,color='w')
ax1.set_ylabel('Chargedoff Proportion',fontsize=14,color = 'w')
plt.show()
# Observations:
# Those who are not working or have less than 1 year of work experience have high chances of getting charged off.
# It makes sense as with less or no experience they don't have source of income to repay loan.
# Rest of the applicants have more or less same chnaces of getting charged off.
In [471]:
# Bivariate Analysis on address state against Chargedoff_Proportion
state_vs_loan = data.groupby(['addr_state', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
state_vs_loan['Total'] = state_vs_loan['Charged Off'] + state_vs_loan['Current'] + state_vs_loan['Fully Paid']
state_vs_loan['Chargedoff_Proportion'] = state_vs_loan['Charged Off'] / state_vs_loan['Total']
state_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# states NE has very high chances of charged off but number of applications are too low to make any decisions.
# NV,CA and FL states shows good number of charged offs in good number of applications.
Out[471]:
| loan_status | addr_state | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 28 | NE | 3.00 | 0.00 | 2.00 | 5.00 | 0.60 |
| 32 | NV | 107.00 | 17.00 | 368.00 | 492.00 | 0.22 |
| 40 | SD | 12.00 | 2.00 | 50.00 | 64.00 | 0.19 |
| 0 | AK | 14.00 | 2.00 | 63.00 | 79.00 | 0.18 |
| 9 | FL | 497.00 | 85.00 | 2252.00 | 2834.00 | 0.18 |
| 24 | MO | 114.00 | 16.00 | 551.00 | 681.00 | 0.17 |
| 13 | ID | 1.00 | 0.00 | 5.00 | 6.00 | 0.17 |
| 31 | NM | 30.00 | 5.00 | 149.00 | 184.00 | 0.16 |
| 11 | HI | 28.00 | 8.00 | 138.00 | 174.00 | 0.16 |
| 4 | CA | 1116.00 | 149.00 | 5767.00 | 7032.00 | 0.16 |
| 36 | OR | 71.00 | 16.00 | 364.00 | 451.00 | 0.16 |
| 43 | UT | 40.00 | 6.00 | 211.00 | 257.00 | 0.16 |
| 20 | MD | 161.00 | 26.00 | 849.00 | 1036.00 | 0.16 |
| 10 | GA | 215.00 | 38.00 | 1135.00 | 1388.00 | 0.15 |
| 46 | WA | 127.00 | 22.00 | 684.00 | 833.00 | 0.15 |
| 30 | NJ | 277.00 | 59.00 | 1490.00 | 1826.00 | 0.15 |
| 29 | NH | 25.00 | 5.00 | 139.00 | 169.00 | 0.15 |
| 27 | NC | 114.00 | 38.00 | 629.00 | 781.00 | 0.15 |
| 22 | MI | 102.00 | 16.00 | 596.00 | 714.00 | 0.14 |
| 39 | SC | 66.00 | 13.00 | 389.00 | 468.00 | 0.14 |
| 3 | AZ | 121.00 | 30.00 | 718.00 | 869.00 | 0.14 |
| 17 | KY | 45.00 | 13.00 | 266.00 | 324.00 | 0.14 |
| 47 | WI | 63.00 | 20.00 | 376.00 | 459.00 | 0.14 |
| 35 | OK | 40.00 | 12.00 | 246.00 | 298.00 | 0.13 |
| 23 | MN | 81.00 | 10.00 | 522.00 | 613.00 | 0.13 |
| 14 | IL | 196.00 | 47.00 | 1262.00 | 1505.00 | 0.13 |
| 33 | NY | 488.00 | 111.00 | 3156.00 | 3755.00 | 0.13 |
| 26 | MT | 11.00 | 2.00 | 72.00 | 85.00 | 0.13 |
| 38 | RI | 25.00 | 4.00 | 165.00 | 194.00 | 0.13 |
| 6 | CT | 94.00 | 24.00 | 618.00 | 736.00 | 0.13 |
| 44 | VA | 176.00 | 38.00 | 1182.00 | 1396.00 | 0.13 |
| 34 | OH | 153.00 | 44.00 | 1020.00 | 1217.00 | 0.13 |
| 5 | CO | 97.00 | 26.00 | 659.00 | 782.00 | 0.12 |
| 18 | LA | 53.00 | 9.00 | 367.00 | 429.00 | 0.12 |
| 1 | AL | 54.00 | 17.00 | 378.00 | 449.00 | 0.12 |
| 48 | WV | 21.00 | 5.00 | 151.00 | 177.00 | 0.12 |
| 19 | MA | 156.00 | 42.00 | 1123.00 | 1321.00 | 0.12 |
| 37 | PA | 178.00 | 49.00 | 1281.00 | 1508.00 | 0.12 |
| 41 | TN | 2.00 | 0.00 | 15.00 | 17.00 | 0.12 |
| 42 | TX | 313.00 | 66.00 | 2315.00 | 2694.00 | 0.12 |
| 16 | KS | 31.00 | 16.00 | 224.00 | 271.00 | 0.11 |
| 45 | VT | 6.00 | 1.00 | 47.00 | 54.00 | 0.11 |
| 2 | AR | 27.00 | 10.00 | 208.00 | 245.00 | 0.11 |
| 25 | MS | 2.00 | 0.00 | 17.00 | 19.00 | 0.11 |
| 8 | DE | 12.00 | 1.00 | 101.00 | 114.00 | 0.11 |
| 7 | DC | 15.00 | 3.00 | 196.00 | 214.00 | 0.07 |
| 49 | WY | 4.00 | 3.00 | 76.00 | 83.00 | 0.05 |
| 21 | ME | 0.00 | 0.00 | 3.00 | 3.00 | 0.00 |
| 15 | IN | 0.00 | 0.00 | 9.00 | 9.00 | 0.00 |
| 12 | IA | 0.00 | 0.00 | 5.00 | 5.00 | 0.00 |
In [472]:
# Drawing bar plots on data calculated above. Try to visualize the pattern to understand the data better.
fig, ax1 = plt.subplots(figsize=(16, 6),facecolor='g')
ax1.set_title('Address State vs Chargedoff Proportion',fontsize=15,color='w')
ax1=sns.barplot(x='addr_state', y='Chargedoff_Proportion', data=state_vs_loan)
ax1.set_xlabel('Address State',fontsize=14,color='w')
ax1.set_ylabel('Chargedoff Proportion',fontsize=14,color = 'w')
plt.show()
# Observations:
# states NE has very high chances of charged off but number of applications are too low to make any decisions.
# NV,CA and FL states shows good number of charged offs in good number of applications.
In [473]:
# Bivariate Analysis on verification_status against Chargedoff_Proportion
verfy_vs_loan = data.groupby(['verification_status', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
verfy_vs_loan['Total'] = verfy_vs_loan['Charged Off'] + verfy_vs_loan['Current'] + verfy_vs_loan['Fully Paid']
verfy_vs_loan['Chargedoff_Proportion'] = verfy_vs_loan['Charged Off'] / verfy_vs_loan['Total']
verfy_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# There is not much difference in charged off proportion.
# This variable doesn't provide any insights for charged off.
Out[473]:
| loan_status | verification_status | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 2 | Verified | 2032 | 593 | 10000 | 12625 | 0.16 |
| 1 | Source Verified | 1422 | 306 | 8135 | 9863 | 0.14 |
| 0 | Not Verified | 2130 | 227 | 14474 | 16831 | 0.13 |
In [474]:
# Drawing bar plots on data calculated above. Try to visualize the pattern to understand the data better.
fig, ax1 = plt.subplots(figsize=(12, 8),facecolor='y')
ax1.set_title('Verification Status vs Chargedoff Proportion',fontsize=15,color='w')
ax1=sns.barplot(x='verification_status', y='Chargedoff_Proportion', data=verfy_vs_loan)
ax1.set_xlabel('Verification Status',fontsize=14,color='w')
ax1.set_ylabel('Chargedoff Proportion',fontsize=14,color = 'w')
plt.show()
# Observations:
# There is not much difference in charged off proportion.
# This variable doesn't provide any insights for charged off.
In [475]:
# Bivariate Analysis on pub_rec_bankruptcies against Chargedoff_Proportion
bankruptcies_vs_loan = data.groupby(['pub_rec_bankruptcies', 'loan_status']).loan_status.count().unstack().fillna(0).reset_index()
bankruptcies_vs_loan['Total']=bankruptcies_vs_loan['Charged Off']+bankruptcies_vs_loan['Current']+bankruptcies_vs_loan['Fully Paid']
bankruptcies_vs_loan['Chargedoff_Proportion'] = bankruptcies_vs_loan['Charged Off'] / bankruptcies_vs_loan['Total']
bankruptcies_vs_loan.sort_values('Chargedoff_Proportion', ascending=False)
# Observations:
# Those who already have pub_rec_bankruptcies value 1, have charged off proportion higher than who have no pub_rec_bankruptcies.
# pub_rec_bankruptcies count 2 has even higher charged off proportion but those numbers are not significant to decide.
# Not known is the column for which we don't have any information abount borrower.
# This also makes sense that who has defaulted before has more chances of dafaulting in future as well.
Out[475]:
| loan_status | pub_rec_bankruptcies | Charged Off | Current | Fully Paid | Total | Chargedoff_Proportion |
|---|---|---|---|---|---|---|
| 2 | 2.00 | 2.00 | 2.00 | 3.00 | 7.00 | 0.29 |
| 1 | 1.00 | 365.00 | 37.00 | 1268.00 | 1670.00 | 0.22 |
| 3 | Not Known | 116.00 | 0.00 | 572.00 | 688.00 | 0.17 |
| 0 | 0.00 | 5101.00 | 1087.00 | 30766.00 | 36954.00 | 0.14 |
In [476]:
# Drawing bar plots on data calculated above. Try to visualize the pattern to understand the data better.
fig, ax1 = plt.subplots(figsize=(12, 7),facecolor='y')
ax1.set_title('Public Bankruptcy Records vs Chargedoff Proportion',fontsize=15,color='w')
ax1=sns.barplot(x='pub_rec_bankruptcies', y='Chargedoff_Proportion', data=bankruptcies_vs_loan)
ax1.set_xlabel('Public Bankruptcy Records',fontsize=14,color='w')
ax1.set_ylabel('Chargedoff Proportion',fontsize=14,color = 'w')
plt.show()
# Observations:
# Those who already have pub_rec_bankruptcies value 1, have charged off proportion higher than who have no pub_rec_bankruptcies.
# pub_rec_bankruptcies count 2 has even higher charged off proportion but those numbers are not significant to decide.
# Not known is the column for which we don't have any information abount borrower.
# This also makes sense that who has defaulted before has more chances of dafaulting in future as well.
Bivariate Analysis : Part 2
In [477]:
# Bivariate Analysis - Puprose of loan vs Loan amount
# Box Plot
plt.figure(figsize=(12,8),facecolor='b')
ax = sns.boxplot(y='purpose', x='loan_amnt', data =data,palette='rainbow')
ax.set_title('Purpose of Loan vs Loan Amount',fontsize=15,color='w')
ax.set_ylabel('Purpose of Loan',fontsize=14,color = 'w')
ax.set_xlabel('Loan Amount',fontsize=14,color = 'w')
plt.show()
# Observations:
# Median,95th percentile,75th percentile of loan amount is highest for loan taken for small business purpose among all purposes.
# Debt consolidation is second and Credit card comes 3rd.
In [478]:
# Bivariate Analysis - Puprose of loan vs Loan amount
# violin Plot - It shows the distribution of quantitative data across several levels of one (or more) categorical variables
# such that those distributions can be compared.
plt.figure(figsize=(14,8),facecolor='b')
ax=sns.violinplot(y='purpose', x='loan_amnt', data =data)
ax.set_title('Purpose of Loan vs Loan Amount',fontsize=15,color='w')
ax.set_ylabel('Purpose of Loan',fontsize=14,color = 'w')
ax.set_xlabel('Loan Amount',fontsize=14,color = 'w')
plt.show()
# Observations:
# Loan taken for small business purpose, Debt consolidation and Credit card are somewhat evenly distributed
# as compare to loan taken for other purposes.
In [479]:
# Bivariate Analysis - Term of loan vs Interest Rate
# Box Plot
plt.figure(figsize=(10,6),facecolor='b')
ax = sns.boxplot(y='int_rate', x='term', data =data,palette='rainbow')
ax.set_title('Term of loan vs Interest Rate',fontsize=15,color='w')
ax.set_ylabel('Interest Rate',fontsize=14,color = 'w')
ax.set_xlabel('Term of loan',fontsize=14,color = 'w')
plt.show()
# Observations:
# It is clear that avearge intrest rate is higher for 60 months loan term.
# Most of the loans issued for longer term had higher intrest rates for repayement.
In [480]:
# Bivariate Analysis - Purpose vs Interest Rate
# Box Plot
plt.figure(figsize=(14,8),facecolor='b')
ax = sns.boxplot(x='int_rate', y='purpose', data =data,palette='rainbow')
ax.set_title('Purpose vs Interest Rate',fontsize=15,color='w')
ax.set_xlabel('Interest Rate',fontsize=14,color = 'w')
ax.set_ylabel('Purpose of Loan',fontsize=14,color = 'w')
plt.show()
# Observations:
# It is clear that avearge intrest rate is highest for small business purpose.
# Loans taken for small business purposes had to repay the loan with more intrest rate as compared to other.
# Debt consolidation is 2nd where borrowers had to pay more interest rate.
In [481]:
# Bivariate Analysis - Grade vs Interest Rate
# Box Plot
plt.figure(figsize=(14,8),facecolor='b')
ax = sns.boxplot(y='int_rate', x='grade', data =data,palette='rainbow',order = 'ABCDEFG')
ax.set_title('Grade vs Interest Rate',fontsize=15,color='w')
ax.set_ylabel('Interest Rate',fontsize=14,color = 'w')
ax.set_xlabel('Grade',fontsize=14,color = 'w')
plt.show()
# Observations:
# A-grade is a top letter grade for a lender to assign to a borrower.
# The higher the borrower's credit grade,the lower the interest rate offered to that borrower on a loan.
# It is clear that intrest rate is increasing with grades moving from A to F.
In [482]:
# Bivariate Analysis - Loan Amount vs Interest Rate
# Box Plot
plt.figure(figsize=(14,8),facecolor='b')
ax = sns.boxplot(y='int_rate', x='loan_amnt_cats', data =data,palette='rainbow')
ax.set_title('Loan amount vs Interest Rate',fontsize=15,color='w')
ax.set_ylabel('Interest Rate',fontsize=14,color = 'w')
ax.set_xlabel('Loan Amount',fontsize=14,color = 'w')
plt.show()
# Observations:
# It is clear that intrest rate is increasing with loan amount increase.
# probably when loan amount is more it is taken for longer loan term, we saw earlier that longer the loan term more the
# interest rate.
In [483]:
# Bivariate Analysis - dti vs Interest Rate
# Box Plot
plt.figure(figsize=(14,8),facecolor='b')
ax = sns.boxplot(y='int_rate', x='dti_cats', data =data,palette='rainbow')
ax.set_title('Debt to income Ratio vs Interest Rate',fontsize=15,color='w')
ax.set_ylabel('Interest Rate',fontsize=14,color = 'w')
ax.set_xlabel('Debt to income Ratio',fontsize=14,color = 'w')
plt.show()
# Observations:
# If your DTI is low enough you may get a lower interest rate.
# Plot shows no significant variation but there is slight increase in interest rate with increase in DTI.
In [484]:
# Bivariate analysis
# BarPlot to show variation of annual income across grade for every loan status.
plt.figure(figsize=(10,6), dpi=80, facecolor='b', edgecolor='k')
ax=sns.barplot(x='grade', y='annual_inc', hue="loan_status", data=data, estimator=np.mean,order='ABCDEFG')
ax.set_title('Grade vs Annual income',fontsize=15,color='w')
ax.set_ylabel('Annual income',fontsize=14,color = 'w')
ax.set_xlabel('Grades',fontsize=14,color = 'w')
plt.show()
# Observations:
# From this we can conclude that the ones getting 'charged off' have lower annual incomes than the ones
# who'paid fully' for each and every grade (i.e. at same interest range)
In [485]:
# Bivariate Analysis - year vs Interest Rate
# Box Plot
plt.figure(figsize=(14,8),facecolor='b')
ax = sns.boxplot(y='int_rate', x='year', data =data,palette='rainbow')
ax.set_title('Year vs Interest Rate',fontsize=15,color='w')
ax.set_ylabel('Interest Rate',fontsize=14,color = 'w')
ax.set_xlabel('Year',fontsize=14,color = 'w')
plt.show()
# Observations:
# Plot shows intrest rate is increasing slowly with increase in year.
Multivariate Analysis - Pair Plots
In [486]:
# Drawing a pair plot for multivariate analysis
loan_pair = data[['loan_amnt', 'annual_inc', 'year','int_rate','loan_status']].copy()
loan_pair.head()
# Take the log of below variables as they are heavily skewed
loan_pair['loan_amnt'] = np.log10(loan_pair['loan_amnt'])
loan_pair['annual_inc'] = np.log10(loan_pair['annual_inc'])
loan_pair['int_rate'] = np.log10(loan_pair['int_rate'])
# Create a pair plot colored by loan status with a density plot of the # diagonal and format the scatter plots.
sns.pairplot(loan_pair, hue = 'loan_status', diag_kind = 'kde',
plot_kws = {'alpha': 0.6, 's': 80, 'edgecolor': 'k'},
size = 4)
plt.show()
# Observation:
# Higher the intrest rate higher charged off ratio
# Higher the annual income higher the loan amount slightly.
# increase in number of charged off with increase in year.
# intrest rate is increasing with loan amount increase
In [487]:
# Above plot is too dense so just pick year 2010 for getting some insights.
sns.pairplot(loan_pair[loan_pair['year'] == 2010],
vars = ['loan_amnt', 'annual_inc', 'int_rate'],
hue = 'loan_status', diag_kind = 'kde',
plot_kws = {'alpha': 0.6, 's': 80, 'edgecolor': 'k'},
size = 4);
# Observation:
# Higher the intrest rate higher charged off ratio
# Higher the annual income higher the loan amount slightly
# Intrest rate is increasing with loan amount increase this results in high charged off.
